On Domain Wall Broadening in Ferroelectric Lithium Niobate and Tantalate

Page 277, AIP Conference Proceedings, Fundamental Physics of Ferroelectrics 2002. On Domain Wall Broadening in Ferroelectric Lithium Niobate and Tantalate Sungwon Kim*, B. Steiner, A. Gruverman, V. Gopalan* * Materials Research Institute and Dept. of Materials Science and Engineering, Pennsylvania State University, University Park, PA 16802 National Institute for Standards and Technology, Gaithersburg, Maryland, 20899 Dept. of Materials Science and Engineering, North Carolina State University, Raleigh, NC 27695 Abstract. We present evidence for broad regions of domain wall strains extending over micrometers in ferroelectric lithium niobate and tantalate due to the presence of cation nonstoichiometry. The influence of a corresponding polarization broadening on intrinsic coercive fields of ferroelectrics is explored. INTRODUCTION The local structure of a ferroelectric domain wall establishes a key link between atomic structure and mesoscale properties of ferroelectrics. This structure has three distinct features: the local polarization gradient, the local strain gradient, and local optical effects. These are interrelated by electrostriction (strain and polarization), elasto-optic effect (strain and optical index), and electro-optic effect (polarization and optical effects). In contrast to ferromagnetic domain walls, where the magnetic moment can rotate across a domain wall from one orientation to another,[1] the strong coupling between ferroelectric polarization and lattice strain restricts the polarization to specific crystallographic directions. As a consequence, while the antiparallel (180 ) ferromagnetic walls can easily be micrometers wide, theoretical first principle calculations in the most important class of oxygen octahedra ferroelectrics show that the antiparallel ferroelectric walls are atomically sharp (order of 0.5 nm).[2] Further, since the lattice polarization is coupled to the lattice strain through electrostriction, the local strain width across such domain walls is expected to be sharp as well. For a commensurate antiparallel domain wall, no optical index changes are expected across such a wall. This picture of an atomically sharp ferroelectric domain wall is a widely accepted and used concept in theoretical formalisms of the local domain wall structure and its motion under external fields. [3] The purpose of this paper is to show that deviations from such a sharp antiparallel domain wall picture can potentially occur in real ferroelectric materials. As a clear evidence, we present X-ray synchrotron studies of ferroelectric domain walls in congruent LiNbO 3 and LiTaO 3 crystals where broad regions of wall strains exist, arising from cation nonstoichiometry. We also present results from piezoelectric scanning force microscopy, which suggest possible wall broadening under external

fields. We conclude by discussing the significance of a broadened wall in lowering intrinsic coercive fields. DOMAIN WALL STRAINS AND CURVATURE The single crystal studied here is a 0.5mm thick z-cut congruent LiNbO 3 single crystal (uniaxial direction normal to the substrate). The crystal composition was congruent, i.e. (Li 0.95 0.04 Nb 0.01 )NbO 3 where the point defects lithium vacancies, Li and niobium antisites, Nb Li are the nonstoichiometric defects.[4] Starting from a single domain state at room temperature, domains of reverse polarity are created by applying electric fields of 210 kv/cm at room temperature as described in detail elsewhere.[5] The nucleation and growth of domains is random, and not controlled in any way, and the final domain configuration is multiple hexagonal domains within the matrix domain state separated by 180 domain walls. The static domain wall analysis reported here were carried out on beamline X23A3 at National Synchrotron Light source at Brookhaven National Labs, NY. This beamline produces a monochromatic X-ray beam with divergence of ~ 1 arc sec (5 10-6 rad). As a result, small regions in the crystal under examination that deviate from the surrounding area either in lattice parameter or in lattice orientation by more than a part per million in the direction of diffraction, do not fulfill the Bragg condition locally, when oriented in the vicinity of the edge of a Bragg diffraction peak. These regions can be imaged at Bragg angle on a high-resolution silver halide film. The lateral resolution in image is determined by the smallest grain size in the silver halide film, which is ~1µm. A series of (0,0,12) Bragg reflection image with 10keV energy is shown in Fig. 1; Bragg angle =30.74. The matrix domain has +z normal to the image plane. The incidence plane in the image is vertical, parallel to the crystallographic y-axis. (a) (b) (c) (d) +y 400µ x FIGURE 1. A series of X-ray synchrotron images of (0,0,-12) reflection of congruent LiNbO 3. Each frame is separated by an incidence angle of 0.003 starting from (a) to (d), and the Bragg peak corresponds to frame(c). Crystallographic directions for matrix domain is shown in frame (d). From Fig. 1(a) to (d), we rotate the sample in steps of 0.003 towards increasing incident angle. Figure 1(c) corresponds to the Bragg peak. The full width of rocking

curve in going from complete destruction of the image on either side of the Bragg angle was 0.018 or ~65 arc seconds. Even when the sample is rotated through the entire rocking curve range from one end to the other, the region inside the hexagonal domains is always brighter than the outside domain by about 20%. This difference in contrast between inside and outside domains arises from different magnitudes of the complex structure factor inside and outside the hexagonal domains. The complex part of the structure factor is believed to arise from X-ray absorption by niobium atoms,[6-8] and gives rise to differences between the structure factor of (hkl) and ( h kl ) reflections. As expected, when the crystal is imaged using (0,0,-12) reflection (the opposite face of the crystal), the contrast was found to reverse from that in Fig. 1, i.e., the matrix domain was observed, on the average, to be brighter than the region inside the hexagonal domains. A clear evidence for wall curvature is seen in Figure 2, which is a magnified image of a small section of frames (a) and (d) in Figure 1. (a) 1. 2. (b) 6. 3. 5. 4. 400µm FIGURE 2. Magnifications of part of images from frame (a) (left) and frame (d) (right) of Figure 1, corresponding to images from either side of the (00.-12) Bragg peak of congruent LiNbO 3. One observes that the set of wall types 1,2,4, and 5 of every hexagonal domain are not parallel to the incidence plane and show an enhanced contrast with a wider region (10-30 µm wide) of associated strain than the walls that are parallel to the incidence plane (walls 3 and 6). This contrast reveals itself more clearly on moving away from the Bragg peak. These images clearly suggest a curvature of lattice planes in the wall region in going from one domain to the other. Such a wall curvature would be expected to cause a deviation from the Bragg angle when it has a finite component along the plane of incidence; this is consistent with the lack of this strong contrast mechanism for walls parallel to the incidence plane in Fig. 1. A closer inspection reveals that the contrast of walls 1 and 2 is opposite to the contrast at the walls 5 and 6, i.e. if one set of walls is bright, the other is dark (Fig. 2(a)), and vice versa (Fig. 2(d)). The projections the incident and diffracted X-ray wavevectors onto the image plane of figures 1 and 2, point in the y direction. Figure 2(a) shows that for incidence angles below the main Bragg peak for the entire crystal, the domain walls 1 and 2 are closer to the diffraction making them bright, as compared to domain walls 5 and 6, which are farther from the Bragg peak, giving them a darker contrast. For

incidence angles higher than the main Bragg peak for the entire crystal, the domain walls 5 and 6 are closer to the peak, and hence brighter, than walls 1 and 2. This provides a clear evidence for curvature of the diffraction planes across the domain walls. This curvature is such that there is a step up in going from the outside matrix domain to the inside of hexagonal domains by crossing perpendicular to a domain wall. We can estimate this step height as follows. Since the main Bragg peak for the bulk of the crystal lies in frame (d) of Fig. 1, and the local Bragg peak for the domain walls lie ~±0.006 from the main Bragg peak, the shear strain at the wall is ±1.05x10-4, which is simply the angular deviation converted to radians and using the small angle approximation for the tangent function. If the average lateral width of the wall curvature is estimated from Figs. 1 and 2 as ~10 µm, then the domain wall step height is ~ 1 nm. This height is also consistent with similar curvature and height reported in antiparallel domain walls in congruent LiTaO 3 crystals using near-field optical microscopy. [9] We now note a few things. First, these broad regions of strains around domain walls appear to be absent in stoichiometric LiNbO 3 and LiTaO 3 crystals, which do not have lithium deficiency. This clearly establishes a strong correlation between the presence of nonstoichiometric point defects and local domain wall strains. Secondly, these strains exist in domain walls created at room temperature in congruent LiNbO 3 and LiTaO 3 by an external field. However, when these crystals are heated to >200 C for even a few minutes, cooled down and observed under synchrotron X-ray beam, these same domain walls are difficult to find.[10] The local wall strains relax and disappear. These features can been explained in the framework of the presence of polar point defect dipoles, such as 4V Li -Nb Li complexes proposed recently. [5] These defect dipoles tend to align with lattice polarization, thereby minimizing the total energy. However, upon domain reversal at room temperature, a corresponding reorientation of these defect dipoles is forbidden due to negligible ionic conductivity, resulting in frustrated dipoles inside the hexagonal domain, for example, in Figs 1 and 2. This transition from frustrated dipoles inside the domains to stable dipoles in the matrix domain appears to result in a strain gradient across the wall that overlays the intrinsic electrostrictive strain at domain walls. A high temperature anneal removes the frustration of dipoles, thereby removing the strain. As a final note, recent studies to be reported elsewhere indicate that the application of an electric field to the crystal results in extending the observed strains to well over 50 microns. PROBING THE POLARIZATION GRADIENT Associated with strain gradients, one would expect polarization gradients as well through the electrostrictive and piezoelectric coupling. Should we therefore expect that small amounts of nonstoichiometry or the application of external fields would also broaden the polarization gradient? If so, how wide is this gradient? Most studies probing width of domain wall (such as electron microscopy) do not directly probe the polarization gradient at a wall. However, one can probe physical quantities such as the piezoelectric coefficient that are directly proportional to the spontaneous polarization value, P s, and attempt to extract this information from the resulting

images through modeling. Here, we present preliminary results on using piezoelectric-mode scanning force microscopy (piezo-sfm) to study domain walls in congruent ferroelectric LiTaO 3, which is isostructural and shows much the same behavior as LiNbO 3. One relevant difference is that domain walls are parallel to the crystallographic x-axes, rather than the y-axes walls in LiNbO 3.[10] The domains are triangular as well, rather than hexagonal. Figure 3 shows a piezo-mode atomic force microscope image of a domain wall in congruent LiTaO 3. The principle of domain imaging in the SFM piezoresponse mode is described in detail elsewhere.[11] Figure 3 also shows a cross-section profile of the piezo-response signal across the domain wall (i.e. along the white line in the image). FIGURE 3. (Left)A piezoelectric-scanning force microscopy image of 180 domain in congruent LiTaO 3, and (Right) the corresponding line-scan across the domain wall in the image. A linear slope in the amplitude of the piezoresponse signal in regions outside the domain boundary is attributed to an artifact due to the image processing to compensate for a sample tilt, which can lead to an apparent broadening. Therefore, we refer not just to the distance between maximum and minimum values of the piezo-response signal (which is ~400nm wide) but also to the actual image of the wall, which was found to be about 120 nm. The best resolution that could be expected can be as small as the radius of the tip-sample contact area, i.e. of the order of several nanometers. However, the observed widths should be treated as upper limits until a detailed modeling of the images is performed, accounting for the physical properties of the sample (thickness, mechanical compliance, dielectric constant, and piezoelectric response), as well as the experimental conditions (scanning rate, tip size, and the inhomogeneous field distribution). If such a detailed simulation of the image is performed, one could extract the polarization width, and any field-induced broadening effects, since the imaging method relies on application of a field to measure piezoelectric displacements.

WALL BROADENING AND COERCIVE FIELDS Coercive fields are the electric fields required for domain reversal and wall motion. Classically, the prediction of this field by Ginzburg-Landau free energy phenomenology is many orders of magnitude higher than experimentally observed fields. For example, in the case of a uniaxial ferroelectric with a second order phase transition, (applicable to LiNbO 3 and LiTaO 3 ), the theoretically predicted coercive field is given by [12] Ps E c ~ ± 0.1925 (1) ε Substituting for the values of P s and ε 33 in Eq. (1), the intrinsic coercive fields are E c ~2750 kv/cm (LiTaO 3 ) and E c ~5420 kv/cm (LiNbO 3 ). These fields reflect the steepest slope of the energy barrier versus polarization that needs to be overcome between the +P s and P s domain states in a double potential well. In contrast, the coercive fields reported in literature for near-stoichiometric LiTaO 3 is ~17 kv/cm, (which is 162 times lower than the theoretical value)[ 5] and for near-stoichiometric LiNbO 3 is ~40 kv/cm (135 times lower than the theoretical value).[13] On comparing with the congruent compositions of these crystals (coercive fields of ~210 kv/cm), the experimental values are ~ 13 times smaller in LiTaO 3 and ~26 times in LiNbO 3. [5,13] This discrepancy is, in general, explained by the presence of preferred nucleation sites for domains. A recent publication has focused on the elimination of domain nucleation sites in a ferroelectric in order to experimentally reconcile the theoretical and experimental values of intrinsic coercive fields.[14] Once a domain nuclei exists, the Miller and Weinreich theory[3] explains the effective lateral motion of pre-existing domain walls, in terms of the probability of overcoming an energy barrier, U, to preferentially nucleate small wedge-shaped domains adjoining the wall, thus effectively advancing the wall laterally. The domain walls in this model are considered atomically sharp. Let us now consider a graded domain wall, given by P = Ps tanh( x / xo ), where x is the wall normal coordinate. If such a wall moves laterally as a whole by one lattice parameter, a, then in the limit of a/x o <<1, then the coercive field contribution of such a system is [12] E c ~ ( Ps ε 33 Clearly, the graded domain wall has a coercive field which is inversely proportional to the wall width, 2x o. This is shown in Figure 4. This effect can be imagined as follows: In the absence if a domain wall, or in the presence of an atomically sharp wall, the polarization at each point has to switch from +P s to P s (or vice versa) and in the process, overcome a large potential barrier defined by the Landau double potential well. On the other hand, if the wall is graded, a small movement of the wall requires only a small change P in polarization, and each part of the wall region has a much smaller energy barrier G to overcome in each step. This sort of millipede approach can advance the wall with lower coercive fields, where the domain wall plays the same role as a dislocation does in mechanical ) 33 a 2x o (2)

deformation of materials. Now, compared with the estimated theoretical coercive fields (Eq. (2)), the coercive field in the presence of domain walls drops by a factor of ~0.385 x o /a, where 2x o is the wall width. For example, for a=0.515 nm (lattice parameters for LiNbO 3 and LiTaO 3 ), the coercive field drops by a factor of ~45 times for a 2x o =120nm wall width. For a wall width of 2x o = 20nm, which is about 40 lattice parameters, the coercive field drops by a factor of 7.5. A precise experimental determination of the polarization wall width is therefore of central importance to the issue of intrinsic coercive fields. If we ignore other mechanisms, and calculate the values of the equivalent domain wall widths, x o required to account for experimentally measured coercive fields in near-stoichiometric lithium niobate (SLN)[13] and lithium tantalate (SLT),[5] we arrive at the upper limits of domain wall widths of x o ~216 nm for near-stoichiometric LiTaO 3 and x o ~181 nm for near-stoichiometric LiNbO 3. For congruent compositions, these widths would be 17.5 nm (LiTaO 3 ) and 34.5nm (LiNbO 3 ). These wall widths are in a similar range as that observed in Fig. 3. FIGURE 4. Calculated coercive field dependence on domain half-wall width, x o, for LiNbO 3 (broken line) and LiTaO 3 (solid line). [Eq. (2)] Inset shows the same curves with an overlay of arrows indicating the positions of experimentally determined coercive fields for congruent lithium niobate (CLN), congruent LiTaO 3 (CLT), and near-stoichiometric LiNbO 3 (SLN) and lithium tantalate (SLT) and the corresponding calculated wall thicknesses to explain their magnitudes solely by this mechanism. There are other contributions to coercive fields that may be equally important that are not considered here. In particular, space charge fields arising from surface reconstructed layers (or dead layer), and equivalent defect fields arising from bulk dipolar point defect complexes that stabilize domains need to be considered as well. However, polarization broadening would appear to be a significant mechanism if present.

CONCLUSIONS The central message of this work is that a possibility exists that local strains and polarization gradients at domain walls in real ferroelectrics may be significantly broadened by the presence of small amounts of defects, or under the influence of external fields. The presence such preexisting broadened domain walls in a ferroelectric will lower the intrinsic coercive fields from their theoretical estimates by a factor of ~0.385 x o /a, where 2x o is the wall width and a is the lattice parameter. Therefore, the precise experimental determination of polarization wall width is of central importance in ferroelectrics. In general, any broad polarization fluctuation in a ferroelectric can locally lower the coercive fields adjacent to the fluctuation. ACKNOWLEDGMENTS We gratefully acknowledge the useful discussions with Dr. Terence Jach and Steven Durbin. This work was supported by the National Science Foundation grants 9984691 and 9988685. REFERENCES 1. Aharoni, A., Introduction to the Theory of ferromagnetism, Oxford, Oxford Science Publication, 2000, pp. 157-182. 2. Padilla, J., Zhong, W., Vanderbilt, D., Phys. Rev. B, 53, R5969-R5973 (1996). 3. Miller, R. C., and Weinreich, G., Phys. Rev., 117, 1460- (1960). 4. Iyi, N., Kitamura, K., Izumi, F., Yamamoto, J. K., Hayashi, T., Asano, H., and Kimura, S., J. Solid State Chem. 101, 340- (1992). 5. Kim, S., Gopalan, V., Kitamura, K., Furukawa, Y., J. Appl. Phys. 90, 2949-2963 (2001). 6. Krausslich, J., Mohrig, H., Kristall und Technik, 9, 811 (1974) 7. Drakopoulos, M., Hu, S. W., Kuznetsov, S., Snigirev, A., Snigireva, I., and Thomas, P. A., J. Phys. D:Appl Phys. 32, A160 (1999). 8. Hu, Z. W., Thomas, P. A., and Webjorn, J., J. Appl. Cryst, 29, 279 (1996). 9. Yang, T. J., Mohideen, U., Phys. Lett. A, 250, 205 (1998). 10. Gopalan, V., Sanford, N.A., Aust, J. A., Kitamura, K., Furukawa, Y., Crystal Growth, Characterization, and Domain Studies in Ferroelectric Lithium Niobate and Lithium Tantalate, in Handbook of Advanced Electronic and Photonic materials and Devices, Vol. 4 Ferroelectric and Dielectrics, edited by H. S. Nalwa, New York: Academic Press, 2000, pp. 57-114. 11.Gruverman, A., Auciello, O., and Tokumoto, H., Annu. Rev. Mater. Sci. 28, 101 (1998); Gruverman, A., Auciello, O., and Tokumoto, H., J. Vac. Sci. Technol. B14, 602 (1996). 12. Kim, S., Gopalan, V., Gruverman, A., Appl Phys. Lett., 80, 2740 (2002). 13. V. Gopalan, T. E. Mitchell, Y. Furukawa, and K. Kitamura, Appl. Phys.Lett. 72, 1981 (1998). 14. S. Ducharme, V. M. Fridkin, A. V. Bune, S. P. Palto, L. M. Blinov, N. N. Petukhova, and S. G. Yudin, Phys. Rev. Lett. 84, 175 (2000); Also see comments by A. M. Bratkovsky and A. P. Levanyuk, Phys. Rev. Lett. 87, 019701-1 (2000), and S. Ducharme, and V. M. Fridkin, Phys. Rev. Lett. 87, 019702-1 (2000).